An Integrated Primary Care Initiative for Child Health in Northern Togo

The ICBHSS initiative uses a community engagement approach to implement a bundle of evidence-based interventions (Table 1), including (1) subsidizing point-of-care user fees for maternal and child health services (consultations, laboratories, and medication costs) at public health centers for pregnant women and children <5^16–19 ; (2) proactive community case management of maternal care and acute pediatric illnesses provided by trained, equipped, supervised, and salaried CHWs^{13,18,20–27} ; (3) clinical mentoring and enhanced supervision by designated clinical mentors assigned to each public sector health center to support quality improvement and adherence to Integrated Management of Childhood Illness guidelines^14,28,29 ; and (4) targeted health facility and supply-chain operational improvements related to basic infrastructure investments and pharmacy supplies.^30–33  In coordination with Togo’s MoH, the initiative was launched in July 2015.

This study was conducted in the catchment areas of 4 public sector health facilities in the Kozah district within the Kara region of northern Togo, including 1 urban site, Adabawere, and 3 rural sites, Djamdé, Kpindi, and Sarakawa. A population-representative cross-sectional household survey was conducted at baseline and annually from 2015 to 2020 among female residents aged 15 to 49. The 2015 baseline survey respondents were restricted to women aged 18 to 49. Data were collected in the same time frame annually (January to February). To facilitate data collection, each facility catchment area was divided into 15 geographic zones by using geographic information system technology, and households were selected from each zone by using a systematic sampling strategy. To avoid bias, data collectors determined collection location starting points by dropping a pen on a map. If >1 female individual resided in a household, participants were chosen by using a Kish selection grid method.³⁴  All participants were consented to participate.

The primary outcome used to determine sample size was under-5 childhood mortality rate per 1000 live births. With a sample size of ∼1500 respondents annually, we were powered to detect a >50% reduction in under-5 mortality³⁵  over the study period in the 4 catchment areas, using a baseline rate of 88 deaths per 1000 live births,⁸  a design effect of 1.5, a total fertility rate of 4.8,⁸  a 20% nonresponse rate, and an α of .05.³⁶ 

Data were collected by using paper surveys (2015) and electronic tablets with questionnaires developed in KoBoToolbox with built in consistency checks (2016–2020) by trained data collectors.³⁷  The questionnaire was adapted from the Demographic Household Survey (DHS) and Multiple Indicator Cluster Survey, respectively implemented by Togo in 2013 and 2010 (Supplemental Fig 3).⁸  Questionnaires were initially developed in English, were then translated into French, and then were back translated into English for accuracy. Data collectors conducted the survey in either French or the local language, Kabiyé, depending on respondent preference. Questions included measures of demographic characteristics, childhood illness prevalence, health service coverage, barriers to care, and health-seeking behavior.³⁸  Programmatic data on process and quality measures were collected at the community and facility level by ICBHSS staff.

Ethical approvals for this study were obtained from the institutional review boards of the Togolese MoH in Lomé, Togo (reference: CBRS/031/2014) and the Albert Einstein College of Medicine, New York (reference: 005127).

Analysis and presentation of data were organized by using a modified Reach, Effectiveness, Adoption, Implementation, Maintenance evaluation framework.³⁹  Frequencies for demographic characteristics and Reach, Effectiveness, Adoption, Implementation, Maintenance measures were reported as raw counts and weighted population percentages, accounting for complex sampling, estimated through the R function svydesign. Weights were created to account for the proportion of households approached in each catchment area, individual household nonresponse, and number of eligible women per household.

Reach measures were defined as the population who received ICBHSS-related interventions related to maternal and child health facility services. These included the proportion of eligible individuals receiving antenatal care during the first trimester, facility-based delivery, postnatal care, and acute care for pediatric illness symptoms.

The primary effectiveness end point was the change in the under-5 mortality rate assessed at baseline and annually. Under-5 mortality was defined as the probability of a child dying between birth and 60 months of age and is expressed as a rate per 1000 live births.⁸  Mortality rates were calculated from household survey data birth tables that asked respondents to provide month and year of birth, survival status of each live birth, and age at death of each deceased live birth occurring in the previous 5 years.⁴⁰  Estimates were calculated from the DHS mortality rates methodology by using a publicly available R package that included population-based survey weights and a Jackknife repeated replication approach to estimate 95% confidence intervals (CIs) (DHS.rates).^41,42  Although this study did not have sufficient power to assess mortality changes in specific time periods after birth (ie, neonatal mortality rate, postneonatal mortality rate, infant mortality rate, or child mortality rate), these rates are also presented to provide additional contextual information.

Adoption was focused on behavior change and was determined by community-based coverage by CHWs, as measured by the proportion of the population (1) reporting a home visit from a CHW, and (2) receiving treatment from a CHW in the last year.

Implementation measures used ICBHSS program data to assess fidelity and were organized by 2 Institute of Medicine quality domains⁴³ : timely and efficient. “Timely” was defined as the time between onset of symptoms and assessment by a CHW or at the health center. “Efficient” referred to the proportion of referrals to higher levels of care that were successful.

Maintenance was assessed by using factors associated with sustainability in sub-Saharan Africa cited in previous studies⁴⁴ : stakeholder involvement, community mobilization, public sector engagement, and cost. Stakeholder and community-level engagement were reported as the number of community forums completed. Public sector engagement was defined as the participation of MoH leadership at the district, regional, and national levels in ICBHSS implementation.

To determine annual maintenance costs, we conducted a retrospective “top-down”⁴⁵  costing analysis using audited financial statements, programmatic data, and financial data for the period between July 1, 2018, and June 30, 2019. All financial data, including direct and indirect costs, were collected from accounting records, whereas programmatic data were captured from mHealth applications and Togo’s Demographic and Health Information System.

All described analyses for reach, effectiveness, and adoption measures were conducted by using R software, version 3.6.3. Analyses were described according to the Standards for Reporting Implementation Studies reporting checklist.⁴⁶ 

From 2015 to 2020, a total of 10 022 households completed surveys (Table 2). Refusal rates were similar across the 4 sites (range 0.3%–1.8%). However, the proportion of residents not at home was notably higher in 2015 (28.4%). There were no meaningful differences observed from 2015 to 2020 in the distribution of demographic or socioeconomic factors (Table 2). The distribution of age group differed in 2015 because of the exclusion of 15- to 17-year-old respondents, but they did not differ from 2016 to 2020.

Before the ICBHSS launch, the under-5 mortality rate was 51.1 (95% CI: 35.5–66.8) deaths per 1000 live births at baseline, with infant mortality (deaths under 1 year) and child mortality (deaths in children aged 1–5 years) contributing approximately equally to the observed rate (Table 3). There was an observed decrease in the under-5 mortality rate from 2015 to 2020 (Fig 1). In year 5 (2020), the under-5 mortality rate was 35.8 (95% CI: 23.4–48.2) deaths per 1000 live births. This reduction in under-5 mortality was accompanied by declines in both the child mortality rate and infant mortality rate. The overall decline in infant mortality was composed of a decline in the postneonatal mortality rate, but no decline was observed in neonatal mortality, with similar rates and CIs reported in 2015 and 2020. Mortality rates are not presented by site because of the small number of events and lack of precision in stratified estimates.

We summarized reach and coverage measures related to maternal and pediatric care as follows (Table 4). The percentage of women reporting receiving any antenatal care in the first trimester of the most recent pregnancy was observed to increase in all sites from 37.5% (95% CI: 33.4%–41.7%) at baseline to 50.1% in year 5 (95% CI: 45.6%–54.7%). In the 3 rural sites, there was an increase in the proportion of the women who reported a health facility–based childbirth, at baseline between 47.6% and 66.0% to 79.5% to 89.4% at year 5 (2020). The proportion of women who received postnatal care after leaving the health facility increased from 38.1% (95% CI: 33.3%–43.1%) at baseline to >60% in each of the last 4 years. Among respondents who sought care for a child with a fever in the 2 weeks preceding the survey, there was an increasing trend of presenting for care within 1 day over the study period, from 51.9% (95% CI: 47.1%–56.6%) at baseline to 80.3% (95% CI: 74.6%–85.0%) at year 5 (2020).

The proportion of the population who reported receiving a CHW home visit in the previous year at baseline, 49.3% (95% CI: 46.7%–51.9%), compared with 58.7% (95% CI: 56.1%–61.1%) at year 5, and treatment at home, 24.1% (95% CI: 21.9%–26.4%) at baseline compared with 45.7% (95% CI 43.3%–48.2%) at year 5 (Fig 2). Analyses disaggregated by site revealed larger gains observed at the 3 rural sites (Djamdé, Kpindi, and Sarakawa), compared with the urban site (Adabawere).

There were 253 290 CHW home visits and 76 388 maternal and pediatric (<5 years old) health facility visits conducted over the 5-year intervention period across the 4 sites (Table 5). Timeliness of treatment, the proportion of cases of child illness treated within 72 hours of onset, improved from 80% in 2016 to 97% in 2020. The proportion of successful referrals to higher levels of care increased overall from 56% in 2016 to 79% in 2019 and 76% in 2020.

As part of the community-based implementation strategy to support stakeholder engagement between 2014 and 2020, the team conducted 1379 community meetings, including town halls, educational events, and exchanges in partnership with MOH staff, CHWs, and community members. From 2016 to 2020, there were 10 program implementation meetings with district and regional health officials and 17 joint supervision visits with district and regional health officials to support public sector integration. We estimated that the total annual cost of the ICBHSS intervention in 2019 was to $389 412, or $8.84 per person per year.

In our study, we observed a decline in the under-5 mortality rate in the Kozah district of Togo after the implementation of the ICBHSS initiative over 5 years. The decline observed in 4 ICBHSS catchment areas was greater, 30%, compared with the estimated 14% decline observed throughout Togo during the same time period, as published by the United Nations Inter-agency Group for Child Mortality Estimation.^6,47  This reduction in mortality was associated with observed increases in health care coverage, use of health services, and improved quality of service delivery known to reduce child mortality. Population-representative community surveys suggested increased in-home care and treatment by CHWs; increased uptake of antenatal care, especially in the first trimester; more births occurring at health facilities; more women receiving postnatal care after leaving the health facility; and improved health-seeking behaviors when children fell ill. In addition, programmatic data suggested improvements in quality. The estimated annual cost of the ICBHSS program was $389 412, or $8.84 per person per year, and would be a significant new investment to the MOH; however, this amount would represent <25% of the estimated per capita health expenditure in Togo ($38 in 2017).⁴⁸  To our knowledge, this is the first pragmatic study in Togo to evaluate a complex evidence-based intervention bundle in terms of both effectiveness and secondary explanatory factors including health-seeking behaviors and clinic use.

The plausibility of observed mortality reduction being in part an effect of the intervention is supported by the “3 delays”^49,50  model, as the ICBHSS initiative was designed to reduce (1) delays in seeking care through CHWs; (2) delays in accessing facility-based care through eliminating of point-of-care costs, improved referrals, and CHWs; and (3) delays in receiving appropriate treatment through health facility and supply-chain operational improvements, clinical mentoring, and enhanced supervision. Our findings are consistent with the evaluation of other evidence-based practice interventions addressing multiple barriers to health care access and quality. Authors of a recent systematic review examining the impact of community-based implementation strategies employing trained health outreach workers (eg, CHWs) found that these programs were associated with a reduction in maternal morbidity, neonatal mortality, stillbirths, perinatal mortality, and likely maternal mortality, with 6 studies from Africa.²²  Improvements in child health and mortality outcomes have also been observed after the implementation of integrated facility- and community-based health systems-strengthening initiatives in Mali,¹⁸  Rwanda,⁵¹  Madagascar,⁵²  the Gambia,²¹  Ghana,⁵³  and Benin.⁵⁴ 

This study has several important limitations. Because we used a pre–post study design without a subnational control group, we cannot rule out larger temporal trends in Togo contributing to the greater-than-expected reductions in under-5 mortality. We are not aware of demographic or socioeconomic changes that could explain the reduction in the district during this time period. Although there are limitations to the comparison with national mortality estimates, the lack of routinely collected reliable mortality data at the subnational level in low-income settings like Togo precludes other assessments. Direct comparison with the United Nations Inter-agency Group for Child Mortality Estimation national estimates is complicated by different methodologic approaches to the estimates. Although collection of full birth histories from household survey data are considered the standard for mortality estimates in low-income countries, there are significant data quality challenges that complicate the ability to monitor yearly changes in mortality through these data.⁵⁵  In addition, it should be noted that estimated mortality rates from this study were based on a small (<100) absolute number of reported child deaths. In addition, although demographic and household surveys are a fundamental data source for assessing reproductive, maternal, and child health care services in low-income countries, issues regarding the validity and associated self-report biases of household survey data have been widely recognized.^35,55  Although we attempted to identify clinic catchment population using our sampling techniques, it is possible that participants surveyed are not the “true” catchment population, that they may seek care at other locations, and that this may be more likely to occur in the urban location. An important limitation includes the distinct number of households “not at home” during baseline sampling, which may have reduced the representativeness of the respondents in 2015. In 2016, because of this high rate, data collectors changed the data collection protocol to include returning to a household 3 times before marking “not at home” to align with Togo national DHS recommendations and Multiple Indicator Cluster Survey guidance. In addition, the age range was changed in 2016 to include 15- to 17-year-olds to similarly align with the Togolese MoH community survey protocols. Additional sensitivity analyses excluding 15- to 17-year-old respondents in the calculation of under-5 mortality in 2020 did not meaningfully change the estimate (not shown). Differences in population density in the urban site compared with the rural sites may have influenced the implementation of the systematic sampling strategy and the representativeness of respondents. In 2018, we initiated a type II pragmatic effectiveness-implementation study including a cluster-randomized stepped-wedge trial to address some of the methodologic limitations cited, specifically adding control clusters, and to further evaluate the impact of the ICBHSS initiative on primary health care uptake and child mortality in northern Togo.⁵⁶ 

We observed gains in access to and quality of evidence-based maternal and child health services that were associated with an observed decline in under-5 mortality over the first 5 years of the ICBHSS initiative. There are several areas of research that remain to better understand this potential causal relationship. The ICBHSS initiative aimed to address a know-do gap related to maternal and child health care for families in Togo and put what is known into practice using community-based implementation strategies. Our experience may provide a path forward to reduce infant and child mortality in similar low-income settings and to eliminate preventable child death globally.

This manuscript is dedicated to our late colleague, Dr Abdourahmane Diparidé Agbèrè, who was a tireless advocate for Togolese children. He was not able to see the results of this work that he supported, but we know that the progress described and the lives of children saved because of his work remain his lasting legacy. We acknowledge and thank the following partners and advisers for their integral assistance in developing and implementing this study: the health authorities in the Kara region and the districts of Kozah; IH staff in Togo and New York; faculty advisers at the Albert Einstein College of Medicine, City University of New York, and the University of Washington; and collaborators from New York University School of Public Health. We extend our gratitude for the partnership between Dr Ari Johnson and his team at Muso and thank Dr Johnson specifically for his help designed the evaluation, review and feedback of this manuscript. We also thank our colleague, Stephanie Anderson, who assisted with the editing process.

CI

confidence interval

CHW

community health worker

DHS

Demographic and Health Survey

ICBHSS

Integrated Community-Based Health Systems-Strengthening

IH

Integrate Health

MoH

Ministry of Health and Public Hygiene